Ownership, Experience and Defects: A Fine-Grained Study of Authorship

Author(s): Foyzur Rahman and Premkumar Devanbu
Venue: Software Engineering (ICSE), 2011 33rd International Conference
Date: 2011

Type of Experiement: Controlled Experiment
Sample Size: 4
Class/Experience Level: Professional
Participant Selection: Large open source software projects between 200,000 to 900,000 LOC
Data Collection Method: Code Metric, Project Artifact(s)


This paper studies how ownership and experience can affect software quality. The authors base this own implicated code, which is code that is modified to fix a defect. The more evidence there is, the less quality of code there is. The authors used the BLAME feature of Git to identify ownership of code on a line by line basis, in order to gain a more fine-grained study of the effects of experience and ownership. They present questions that ask how related implicated code is in relation to the number of contributors to the code, whether general or more specialized experience is associated with implicated code, and whether the amount of ownership a single person has to a field is associated with implicated code.

The authors used 4 medium to large-sized (from 200,000 to 900,000 LOC) open source projects to study the effects of ownership and experience, and submitted them to Git repositories in used Git Blame in order to gain data on a line by line basis.

Results showed that implicated code is less likely to involve contributions from multiple developers, and that it is mostly contributed by a single author. In addition, it shows that general or specialized experience of the author has no clear association with implicated code. In the case of the number of contributors to code, the authors cite several other papers, such as Weyuker and Graves, that show the number of contributors was not a factor, and therefore suggests that number of contributors has a more complex role to defect-proneness.